JP2006507579A - Histological evaluation of nuclear polymorphism - Google Patents

Abstract

A method of histological assessment of nuclear pleomorphism to identify potential cell nuclei divides image data into overlapping sub-images. It uses principal component analysis to derive monochromatic image data, followed by Otsu thresholding to produce a binary image. It removes image regions at sub-image boundaries, unsuitably small image regions and holes in relatively large image regions. It then reassembles the resulting sub-images into a single image. Perimeters (P) and areas (A) of image regions which are potential cell nuclei are determined and used in calculating nuclear shape factors P2/A. Nuclear pleomorphism is assessed as relatively low, moderate or high according to whether predetermined shape factor thresholds indicate a mean cell nucleus shape factor for an image is relatively low, moderate or high.

Description

  The present invention relates to a method, apparatus and computer program for histological evaluation of nuclear polymorphism, and in particular to (but not limited to) histological image evaluation that provides clinical information regarding cancer such as breast cancer. The method of the present invention is also relevant not only to breast cancer but also to other cancers such as colon cancer and cervical cancer.

  Breast cancer is a common type of cancer in women and occurs to a lesser extent in men. When a lesion indicative of breast cancer is detected, a tissue sample is taken and examined by a histopathologist to plan for diagnosis, prognosis and treatment. However, pathological analysis of tissue samples is time consuming and a subjective technique. This analysis involves interpretation of an image by the human eye, and has the feature that even if the time is different even by the same observer, even if the same sample is observed, they do not agree. For example, two different observers that evaluate images of the same tissue sample may have different views on multiple images. Opinions about as much as 30% of images may vary. The problem is exacerbated by heterogeneity, i.e., by the complexity of the characteristics of a particular tissue sample.

  There is a need to support pathologists' diagnosis and patient treatment by realizing objective measurements of the degree of nuclear polymorphism.

  The present invention realizes a method of histological evaluation of nuclear polymorphism by specifying an image region potentially corresponding to a cell nucleus of tissue image data, and this method binarizes image data. Threshold processing, determining the peripheral length and area of the identified image region, calculating the shape factor of the image region from the peripheral length and area, and statistical of the shape factor Evaluating polymorphism from the parameters.

  An advantage of the present invention is that it provides information for pathologist diagnosis and patient treatment by providing objective measurements of polymorphism.

  The statistical parameters of the form factor can consist of at least one of their average, weighted average, median, mode, maximum and minimum. The step of thresholding the imaged data may be an Otsu threshold processing method.

によって与えられる。ここで、ｋは定数、Ｐは画像領域の周辺長、Ａは画像領域の面積であり、細胞核に潜在的に対応する一連の画像領域に対する平均形状因子Ｓ_ｍは、Ｓ_ｍ≦３０ｋ（低い）、３０ｋ＜Ｓ_ｍ≦３５ｋ（中間）およびＳ_ｍ＞３５ｋ（高い）として区分でき、それぞれ相対的に低い、中間、または高い多形性に対応する。 The step of assessing polymorphism determines the polymorphism as relatively low, medium or high depending on whether the mean or median of the form factor is relatively low, medium or high, respectively. be able to. The shape factor S of the image area potentially corresponding to the cell nucleus is

Given by. Here, k is a constant, P is the peripheral length of the image region, A is the area of the image region, and the average shape factor S _m for a series of image regions potentially corresponding to cell nuclei is S _m ≦ 30k (low) , 30k <S _m ≦ 35k (intermediate) and S _m > 35k (high), each corresponding to a relatively low, intermediate, or high polymorphism.

  Prior to the thresholding and binarization of the image data, the color image data is converted to a grayscale image using an improved image definition compared to the individual red, green or blue planes of the color image data. The step of converting to data and then thresholding the image data is performed on the grayscale image data. The step of converting color image data into grayscale image data can be performed by principal component analysis (PCA), in which grayscale image data is the first principal component.

  The step of identifying image regions that potentially correspond to cell nuclei in tissue image data includes filtering of the image data and includes regions that are not targeted using a filtering process that does not appreciably affect the periphery of the image region. Overwrite. The overwriting step can include setting a relatively small image region to a background pixel value and setting a relatively large image region hole pixel to a pixel value of a non-hole image region.

Image regions that potentially correspond to cell nuclei in tissue image data can be identified by a procedure that includes: That is,
a) dividing the image data into overlapping sub-images;
b) applying PCA to each sub-image to generate a respective grayscale sub-image;
c) From the grayscale sub-image,
i) an image area that touches or intersects the sub-image boundary;
ii) an inappropriate small image area;
iii) removing holes in a relatively large image area;
d) Reconstructing the sub-image into a binary image.

  In another aspect, the present invention provides an apparatus for histological assessment of nuclear polymorphism by identifying image regions that potentially correspond to cell nuclei in tissue image data, the apparatus being programmed. By incorporating a computer, the image data is thresholded and binarized, the peripheral length and area of the specified image region are determined, the shape factor of the image region is calculated from the peripheral length and area, It is characterized by evaluating polymorphism from statistical parameters of form factors.

  In another aspect, the present invention is a computer having instructions for controlling a computer to identify an image region that is used in histological assessment of nuclear polymorphism and that potentially corresponds to a cell nucleus of tissue image data. The software implements and binarizes the image data by threshold processing, determines the peripheral length and area of the specified image region, and determines the shape factor of the image region from the peripheral length and area. It is characterized by having instructions for controlling the computer to calculate and evaluate polymorphism from the statistical parameters of the form factor.

  The computer apparatus and computer software aspects of the invention may have preferred features equivalent to the corresponding method aspects of the invention.

  For a more complete understanding of the present invention, embodiments of the invention are described below, by way of example only, with reference to the accompanying drawings.

  FIG. 1 shows a procedure 10 for evaluating a tissue sample in the form of a histopathology slide of breast cancer. This figure illustrates the process of measuring the degree of polymorphism used to assess patient status.

  Procedure 10 uses a database 12 that holds digitized image data obtained from tissue slides, as described below. A section specimen is removed (cut) from a breast tissue sample (living tissue) and placed on each slide. Slides are stained with the stains hemotoxylin and eosin (H & E). H & E is a staining solution that is widely used to delineate tissue and cellular structures. Polymorphism is assessed using tissues stained with H & E.

  Nuclear polymorphism measures the degree of shape variability of the nuclei of tumor cells within a tissue. In normal tissues, cell nuclei have a regular structure in terms of shape and size, but cancer cell nuclei have larger and irregular shapes, with significant variations in shape and size.

  In the prior art manual operation, the clinician places the slide under a 40 × magnification microscope and examines the polymorphism indication in that area (often referred to as a tile). This manual operation requires a pathologist to subjectively evaluate the abnormalities in the size and shape of the cell nuclei of the tissue sample. The values obtained in this way are combined to produce a single measurement result for use in diagnosis. The process of the present invention replaces the subjective procedure of the prior art with an objective procedure.

  In an embodiment of the present invention, image data was obtained from a tissue slide by a pathologist using a Zeiss Axioskop microscope with a Jenoptiks Progress 3030 digital camera. The image data from each slide is a series of digital images obtained with a length magnification of 40 times (i.e., length 40 times, area 1600 times), and each image is electronically equivalent to a tile. Images obtained with other microscopes and cameras may be used.

  To obtain an image, the pathologist scans the entire slide with a microscope and selects the area of the slide that appears to be the most effective in terms of polymorphism assessment at 40x magnification. Each of these areas is then photographed using the microscope and camera described above, thereby producing a digitized image of each of the three colors, red, green and blue (R, G & B). . An image is generated as a combination of the R, G, and B image planes by obtaining three intensity values for each pixel in the pixel array. This image is temporarily stored at 12 for later use. The polymorphism measurement by process 16 requires two tiles. The result of process 16 is converted to a measurement at 20 and becomes the input of a diagnostic report at 22.

  Referring now to FIG. 2, the polymorphic feature detection process 16 is shown in more detail. The polymorphism detection process 16 is performed for each of the two tiles or digitized images described above (raw (RGB) input images). One such image will be described. In a first stage 30, the raw (RGB) input image is divided into overlapping windows of 128 × 128 pixel size. In each window, 64 pixels overlap with 64 pixels in adjacent windows in both horizontal and vertical directions. For example, a 256 × 256 image gives 3 × 3 sets of 128 × 128 overlapping windows. Thus, each window overlaps with each half of its row and column adjacent windows. A technique called “principal component analysis” (PCA, Karhunen-Loeve Transform) is applied to each window. PCA is a Jolly I. T.A. This is a standard mathematical technique described in “Principal Component Analysis” by Springer series in statics, Springer Verlag, 1986. This is “for the principal component” by Jackson JE. It is also described in “User Guide to Principal Components” (John Wiley & Sons, 1991, pp1-25).

  PCA is a transformation into a domain or display where data is more easily utilized or understood (eg, separable or classified into different classes or categories). PCA is a standard mathematical technique for converting a series of correlated variables (as much as possible) into a smaller number of uncorrelated variables called principal components. A principal component is associated with each eigenvalue. The first principal component has the largest eigenvalue and causes as much variable variability as possible compared to the other principal components. In this first principal component, the image definition, contrast and properties are improved compared to the R, G or B image plane of the color image data. Only the first principal component was used in this example, but in another method, if their fixed values are not significantly different, a linear combination of all principal components may be used and the linear combination is appropriate. There is also.

  There are other transforms (analytical methods or filters) that can be used as well and that have specific characteristics (eg, linear or non-linear, utilization data characteristics, mean, variance, etc. in various ways). Such transformations can also be performed using neural / adaptive circuitry and related techniques.

PCA involves calculating a covariance matrix and determining its fixed values and eigenvectors. Here, the image is treated as having an N × 3 matrix, that is, having N pixels and three planes (R, G and B). The covariance matrix is calculated for the matrix element C _{i, j} using the following equation:

Here, C _{i, j} is the covariance of variable i with respect to variable j, x _k and y _k are the i th and j th feature values of the k th object, and μ _x is all N of x _k. Is the average of the values, and μ _y is the average of all N values of y _k . The variance matrix is 3x3, and PCA derives three eigenvectors. The eigenvector is treated as a 3 × 3 matrix, and this matrix is used to multiply a transposed matrix of an N × 3 image matrix to generate a product matrix. The product matrix has N × 1 first columns as the first principal component, and can be regarded as the most important component. The product matrix is the component with the largest eigenvalue and provides the grayscale sub-image (one pixel value for each of the N pixels) with the largest range of information compared to the same value corresponding to the other components. Accordingly, the PCA generates monochrome or gray scale image data. PCA is performed on each of the overlapping windows defined above, each generating a first principal component and a grayscale sub-image of 128 × 128 pixel size.

  At 32, a thresholding technique called “Otsu” is applied to each sub-image obtained from 30 to convert each sub-image into a binary sub-image. Otsu is Otsu N. "A thresholding selection from gray level histogram" by IEEE Trans Systems, Man & Cybernetics, vol. 9, 1979, pp 62-66. Threshold processing technique. The Otsu threshold selection method aims to minimize the ratio of intraclass variance to interclass variance for two classes. That is, the greater the variance between classes, the greater the separation. The Otsu technique is a particularly preferred threshold processing method. In the embodiment of the present invention, the two classes are the sub-threshold class (pixel value 0) and the over-threshold class (pixel value 1). Therefore, the Otsu threshold processing is performed for each gray scale. The sub-image is converted into a binary sub-image that includes a series of blobs. Here, a blob is an image region (an object in the image), and each of the image regions is a group of pixels that are in contact with each other and all have a value of 1 and are considered to correspond to cell nuclei. Blobs may have holes (pixel value 0) in them.

  At 34, all blobs that touch or intersect the sub-image boundary are removed. Thus, if a blob touches the boundary at any pixel, the blob is removed by setting that pixel to the background pixel value. This is because the boundary line in contact with the blob produces an apparent straight blob edge that can result in a later misunderstanding. Because sub-images overlap, blobs that partially appear in one image may appear entirely in another sub-image and are therefore not necessarily lost.

  At 36, the sub-image from 34 is inverted (pixel value 0 changes to 1 and vice versa) and connected component labeling (CCL) is applied to remove the blob holes. CCL is a Klette R.C. And Zamperoniu P. et al. "Handbook of Image Processing Operators" by John Wiley & Sons, 1996, and Rosenfeld A. et al. And Kak A. C. (Digital Picture Processing) (vols. 1 & 2, Academic Press, New York, 1982), a known image processing technique (sometimes referred to as “blob coloring”). The CCL gives a label to each group of consecutive pixels having a pixel value of 1. For reasons of image inversion, the area of pixel value 1 labeled with CCL is here a hole in the blob with the background pixels. The holes in each blob are then removed (filled) by setting those pixels to the values of the other pixels in the blob. The background pixels remain unchanged.

  At 38, the sub-image from 36 is inverted again and CCL is applied again. Due to this second inversion, the area labeled next with CCL becomes the filled blob in each sub-image. The CCL has a function of removing blobs having an area smaller than a user-defined minimum area threshold. In this example, if a 40 × magnification is used, the minimum area threshold in this example is 400 pixels. Thus, all blobs having an area of less than 400 pixels are removed and assimilated into the background by setting those pixels to the background value (0). The remaining blobs with a minimum area of 400 pixels are accepted for further processing as a series of labeled blobs. The CCL further provides a perimeter length P and an area A for a number of pixels in each case for each remaining blob.

  After steps 30 to 38, each sub-image removes unnecessary small blobs and all remaining blobs are filled and the holes in them are removed, so that all the remaining blobs are all the same value pixels. Will be composed. The advantage of steps 30-38 is that they provide spatial filtering but do not appreciably affect the peripheral shape of the blob. This is important for subsequent processing. Such filtering is not essential, but helps reduce the processing burden.

  At 40, the sub-image output from step 38 is reconstructed into a new binary image. The new binary image has the same size as the original raw (RGB) input image and contains only blobs that are evaluated in a subsequent polymorphic process.

は、各ブロブに対して、３８においてＣＣＬで得られたブロブの周辺長Ｐと面積Ａから計算される。適切であれば、Ｓは

の倍数または分数（ｆｒａｃｔｉｏｎ）であってもよく、ＰおよびＡの他の関数を用いて、形状を特定することもできる。Ｓの値はブロブ形状の不規則性が増すにつれて増加し、完全な円形に対しては４π（〜１２．５７）である。Ｓの平均値Ｓ_ｍは、画像の全ブロブのＳ値を合わせて加えて、それらの数で割ることによって完全な画像に対して計算される。Ｓ_ｍをしきい値処理して、オリジナルの生の（ＲＧＢ）入力画像の多形性の測定値を導き出す。Ｓ_ｍしきい値は２００の多形性画像のテスト群の分析から導き出された。以下で表すように、多形性スコア１、２および３のそれぞれに対して３つの区分、Ｓ_ｍ≦３０（低い）、３０＜Ｓ_ｍ≦３５（中間）およびＳ_ｍ＞３５（高い）がある。これらのしきい値は

に対してであって、他の形状因子表現は別のしきい値を必要とすることもある。 Steps 30 through 40 are pre-processing steps that identify a series of blobs in the original raw (RGB) input image. Each of these blobs should now correspond to the cell nucleus. At 42, statistical analysis is applied. Form factor

Is calculated for each blob from the peripheral length P and area A of the blob obtained in CCL at 38. If appropriate, S

Multiples or fractions, and other functions of P and A can be used to identify the shape. The value of S increases with increasing blob shape irregularity and is 4π (˜12.57) for a perfect circle. The average value S _m of S is calculated for the complete image by adding the S values of all blobs of the image together and dividing by their number. S _m is thresholded to derive a measure of polymorphism of the original raw (RGB) input image. The _Sm threshold was derived from analysis of a test group of 200 polymorphic images. As shown below, for each of the polymorphic scores 1, 2 and 3, there are three categories, S _m ≦ 30 (low), 30 <S _m ≦ 35 (middle) and S _m > 35 (high). is there. These thresholds are

Other form factor representations may require different thresholds.

においてｋは一定であり、３つの区分はＳ_ｍ≦３０ｋ（低い）、３０ｋ＜Ｓ_ｍ≦３５ｋ（中間）およびＳ_ｍ＞３５ｋ（高い）となる。

K is constant, and the three sections are S _m ≦ 30 k (low), 30 k <S _m ≦ 35 k (middle) and S _m > 35 k (high).

S _m can also be calculated by another procedure. S _m may simply be the median of an odd number of S values or the average of two median values of an even number of S values. For S _m , a mode value can be used as long as it is a distribution with the maximum value or the minimum value of a series of values of S, or with a substantially single mode value of S values. However, this affects the threshold. The weighted average S _wm can be calculated by multiplying each value of S by the weighting factor W _i , adding the weight values, and dividing their sum by the sum of the weights. Each weight represents the probability of the associated blob or cell nucleus and gives a reliable indication of polymorphism. That is,

  The above description shows that there are multiple ways to determine the overall form factor from a set of statistical characteristics of the form factor, for example, their average, weighted average, median, mode, maximum, and All minimum values indicate that they can be used individually.

At 20, the score measuring nuclear polymorphism has the value 1, 2 or 3, and each score is obtained for each input tile. The maximum of these two scores is taken as the overall polymorphic score.

  Nuclear polymorphism measurements are combined with other measurements obtained for mitosis and tubules using a variety of methods, and the overall rating referred to in medicine as “Bloom and Richardson grading” Or get a modified version of this rating. This result is used by clinicians as a measure of cancer status.

  The examples given in the above description of the calculation results can be clearly evaluated by the appropriate computer program for the carrier medium and execution on a conventional computer system. Since such a program has a known procedure, it is easy for an experienced programmer to implement without requiring the present invention, and will not be described in detail.

FIG. 5 is a block diagram of a procedure for measuring polymorphism to assist in diagnosis and treatment policy decisions. FIG. 2 is a block diagram illustrating in more detail the polymorphic feature detection used in the procedure of FIG. 1 according to the present invention.

Claims (29)

  A method for histologically evaluating nuclear polymorphism by identifying an image region potentially corresponding to a cell nucleus in tissue image data, the image data being thresholded and binarized; Determining the perimeter length and area of the identified image region, calculating the image region shape factor from the perimeter length and area, and evaluating polymorphism from the statistical parameters of the shape factor A method characterized by that.   The method of claim 1, wherein the statistical parameters of the form factor comprise at least one of their average, weighted average, median, mode, maximum and minimum values.   The method of claim 1, wherein the thresholding of the imaged data is an Otsu thresholding method.   The step of assessing polymorphism is to determine that polymorphism is relatively low, medium or high, depending on whether the mean or median form factor is relatively low, medium or high, respectively. The method of claim 1, wherein: The shape factor S of the image area potentially corresponding to the cell nucleus is

(K is a constant, P is an image region peripheral length, and A is an image region area), and the average shape factor S _m for a series of image regions potentially corresponding to cell nuclei is relatively low or intermediate, respectively. Or according to high polymorphism, classified as S _m ≦ 30k (low), 30k <S _m ≦ 35k (intermediate) and S _m > 35k (high). Method.   Prior to the thresholding and binarization of the image data, the color image data is converted to grayscale image data using an improved image definition compared to the individual red, green or blue planes in the color image data. The method of claim 1, wherein the step of converting to: is performed and the step of thresholding the image data is performed on grayscale image data.   7. The method of claim 6, wherein the step of converting color image data to grayscale image data is performed by principal component analysis (PCA), wherein the grayscale image data is a first principal component.   The step of converting the color image data to grayscale image data is to decompose the color image data into a series of sub-images, each overlapping with a respective half of the adjacent image, and PCA is performed on each sub-image. The method of claim 7, further comprising removing a sub-image edge from each sub-image region potentially corresponding to a cell nucleus.   The step of identifying image regions that potentially correspond to cell nuclei in tissue image data is not targeted by filtering the image data using a filtering process that does not appreciably affect the perimeter of the image region. The method of claim 1, comprising overwriting the region.   Overwriting the non-target area includes setting a relatively small image area to the background pixel value and setting a hole pixel in the relatively large image area to a pixel value of the image area without a hole. The method according to claim 9, wherein: Identifying image regions that potentially correspond to cell nuclei in tissue image data involves dividing the image data into overlapping sub-images and applying PCA to each sub-image to generate a respective gray-scale sub-image. And from a grayscale sub-image,
a) an image area that touches or intersects the sub-image boundary;
b) an inappropriately small image area;
c) removing holes in a relatively large image area;
The method of claim 1, comprising reconstructing a sub-image into a binary image.   An apparatus for histologically evaluating nuclear polymorphism by identifying an image region that potentially corresponds to a cell nucleus in tissue image data, the apparatus thresholding the image data to perform binary processing Equipped with a computer programmed to determine the perimeter length and area of the identified image region, calculate the image region shape factor from the perimeter length and area, and evaluate polymorphism from the statistical parameters of the shape factor A device, characterized in that   The apparatus according to claim 12, characterized in that the statistical parameters of the form factors comprise at least one of their average, weighted average, median, mode, maximum and minimum values.   13. The apparatus of claim 12, wherein the computer is programmed to threshold image data using Otsu threshold processing.   The computer is programmed to determine the polymorphism as relatively low, medium or high, depending on whether the mean or median form factor is relatively low, medium or high, respectively. The device according to claim 12, characterized in that: The computer calculates the shape factor S of the image area potentially corresponding to the cell nucleus,

(K is a constant, P is the image region perimeter, A is the image region area), and the computer further determines an average shape factor S _m for a series of image regions potentially corresponding to the cell nucleus, Classify the average form factor as S _m ≦ 30k (low), 30k <S _m ≦ 35k (middle) and S _m > 35k (high) and display polymorphism as relatively low, medium or high, respectively The device according to claim 15, wherein the device is programmed.   Prior to thresholding and binarizing the image data, the computer grayscales the color image data using an improved image definition compared to the individual red, green or blue planes in the color image data. The apparatus of claim 12, wherein the apparatus is programmed to convert to image data, and wherein the computer is programmed to perform thresholding of the image data using grayscale image data.   18. The apparatus of claim 17, wherein the computer is programmed to convert color image data to grayscale image data using a PCA in which grayscale image data is the first principal component. . Computer
a) dividing the image data into overlapping sub-images;
b) applying PCA to each sub-image to generate a gray-scale sub-image,
c) From the grayscale sub-image,
i) an image area that touches or intersects the sub-image boundary;
ii) an inappropriate small image area;
iii) remove the holes in the relatively large image area,
The apparatus according to claim 18, wherein the apparatus is programmed to reconstruct a sub-image into a binarized image.   The computer is programmed to set a relatively small image area to a background pixel value and to set a hole pixel in a relatively large image area to a pixel value of an image area without a hole. The apparatus according to claim 12.   Computer software for use in histological evaluation of nuclear polymorphism and having instructions for controlling a computer to identify image regions that potentially correspond to cell nuclei in tissue image data, the software further comprising: , Binarize the image data by threshold processing, determine the perimeter length and area of the specified image area, calculate the image area shape factor from the perimeter length and area, and polymorph from the shape factor statistical parameters Computer software comprising instructions for controlling a computer to evaluate sex.   The computer software of claim 21, wherein the statistical parameters of the form factor comprise at least one of their average, weighted average, median, mode, maximum and minimum values.   22. Computer software according to claim 21, comprising instructions for controlling the computer to threshold data imaged using the Otsu thresholding method.   Has instructions to control the computer to determine that the polymorphism is relatively low, medium or high, depending on whether the mean or median of the form factor is relatively low, medium or high, respectively The computer software of claim 21, wherein: a)

Determine the shape factor S for the image region potentially corresponding to the cell nucleus given by (k is a constant, P is the image region perimeter, A is the image region area),
b) dividing the mean form factor S _m for image regions potentially corresponding to cell nuclei as S _m ≦ 30k (low), 30k <S _m ≦ 35k (middle) and S _m > 35k (high);
25. Computer software according to claim 24, comprising instructions to control the computer to c) display polymorphism as relatively low, medium or high respectively.   Prior to thresholding and binarizing the image data, the computer grayscales the color image data using an improved image definition compared to the individual red, green or blue planes in the color image data. 24. The method of claim 21, further comprising instructions for controlling the computer to convert to image data and then the computer performs thresholding of the image data using the grayscale image data. Computer software described in   27. Computer software according to claim 26, comprising instructions for controlling the computer to convert color image data into grayscale image data by PCA, wherein the grayscale image data is the first principal component. a) converting the color image data into grayscale image data by separating the color image data into a series of sub-images, each overlapping with a respective half of the adjacent image;
b) Perform PCA on each sub-image,
c) From each sub-image,
i) an image area that touches or intersects the sub-image boundary;
ii) an inappropriately small image area;
iii) remove the holes in the relatively large image area,
28. Computer software according to claim 27, characterized in that it has instructions to control the computer to d) reconstruct the sub-image into a binarized image.   The method of claim 21, further comprising: an instruction for setting a relatively small image area as a background pixel value and setting a hole pixel in a relatively large image area as a pixel value of an image area without a hole. Computer software.

Images